HUMAN DISTURBANCE AND THE HAULING OUT BEHAVIOUR OF STELLER SEA LIONS (Eumetopias jubatus)

HUMAN DISTURBANCE AND THE HAULING OUT BEHAVIOUR OF STELLER SEA LIONS (Eumetopias jubatus) by Laura Kucey B.A. Environmental Biology and Geography Colgate University, 2000 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES Department of Zoology THE UNIVERSITY OF BRITISH COLUMBIA May 2005 Laura Kucey, 2005

Abstract There is considerable interest in assessing and mitigating disruptive effects of humans on the behaviour of marine mammals, especially for species with uncertain or decreasing population trends. Steller sea lions (Eumetopias jubatus) have been under intensive study throughout their range over the past few decades in an attempt to identify the causes of a large population decline in the Gulf of Alaska and Aleutian Islands. Consequently, disturbance due to scientific research has also increased at rookeries and haulouts. The purpose of my study was to determine if there were measurable short-term effects of human disturbance on the numbers of Steller sea lions using terrestrial sites. Numbers and composition of sea lions were documented for 2 3 week periods in southeast Alaska and British Columbia during summer (n = 8 sites) and winter / spring (n = 6 sites). They revealed considerable daily variation in numbers of sea lions hauled out within and among study sites that was related in part to prevailing environmental conditions. However, counts could not be corrected to account for environmental influences on the total numbers of sea lions using haulouts. Hauling out trends were examined for pre- and post-disturbance periods across multiple sites over two seasons. Predetermined research disturbances occurred to collect scats at the haulouts, and to brand pups at the rookery. Three methods were explored to assess local population recovery that addressed both quantitative and temporal aspects of sea lions returning to the study locations. Disturbances resulted in significantly fewer sea lions using haulouts during the post-disturbance period. Variation in the numbers of animals using the haulouts increased following the disturbance, but rates of change in daily numbers did not differ significantly between periods. Six of ten disturbed sites reached full recovery (100% of the pre-disturbance mean) on average 4.3 days after the research disturbance. To determine if individual behaviour was affected by disturbance, sea lions arriving on shore were followed to determine normal patterns of interactions and behaviour. Significant differences were noted in hauling out behaviour between animals that remained on land and those that returned to the water. Sea lions that returned to the water exhibited higher rates of behaviour and interactions with other animals during the week that followed the disturbance. Seasonal differences were also noted in the rates of behaviour and interactions ii

that may be indicative of certain times of the year when sea lions are more sensitive to human presence and disturbance. Increasing levels of human sea lion contact are expected as more and more people visit the remote coastal habitat of Steller sea lions. Future studies are needed to assess the influence of disturbance on sea lion redistribution within a critical recovery period, as well as to determine the physiological effects that sea lions experience with repeated human disturbance. Disturbance studies are an important aspect of conservation initiatives because they can help guide policies and establish restrictions to protect wild populations from human intrusion. iii

Table of Contents Abstract...ii Table of Contents...iv List of Tables...vi List of Figures...vii Acknowledgments... viii Chapter 1: General Introduction: Human disturbance and the hauling out behaviour of Steller sea lions...1 Overview...3 Literature cited...3 Chapter 2: Single census counts may not accurately reflect numbers and trends of Steller sea lions using haulout sites...5 Summary...5 Introduction...5 Methods...7 Data collection...7 Generalized linear models...9 Results...10 Numbers and variation of sea lions on land...11 Age / sex composition...11 Factors that affect numbers of sea lions on land...13 Discussion...15 Numbers of sea lions on land...16 Environmental influences...16 Estimated variation...18 Conclusions...19 Literature cited...20 Chapter 3: Major human disturbances cause short-term reductions in numbers of hauled out Steller sea lions...23 Summary...23 Introduction...23 Methods...26 Results...28 Recovery...31 Movement to and from haulouts...33 Non-research disturbances...35 Discussion...37 Variation in numbers of sea lions on land...37 Steller sea lions and disturbance...38 Recovery...41 Effects of disturbance...42 Future research...43 Conclusions...44 Literature cited...45 iv

Chapter 4: Disturbance does not appear to significantly affect the hauling out behaviour of individual Steller sea lions...48 Summary...48 Introduction...48 Methods...49 Cox proportional hazards model...51 Generalized linear models...52 Results...52 Likelihood of a sea lion remaining on land after hauling out...52 Hauling out behaviour...54 Rates of sea lion interactions...56 Discussion...57 Pinniped terrestrial behaviour...57 Hauling out...58 Behavioural differences...58 Effects of disturbance...59 Literature cited...60 Chapter 5: Concluding Chapter...62 Factors that influence haulout numbers... 62 Population level effects of disturbance... 63 Individual level effects of disturbance... 64 Strengths and weaknesses... 64 Implications... 65 Literature cited... 67 v

List of Tables Table 2.1. Coefficients of variation for all count data and midday counts... 12 Table 2.2. Maximum percent difference in daily counts... 12 Table 2.3. Factors affecting numbers of sea lions hauled out... 14 Table 3.1. Sea lion haulout recovery...31 Table 3.2. Point count haulout recovery... 32 Table 3.3. Daily mean haulout recovery... 33 Table 3.4. Study site details (latitude and longitude, distance to closest alternative haulout)... 34 Table 4.1. Factors influencing the rates of behaviours and sea lion interactions... 54 vi

List of Figures Figure 2.1. Study site locations and proportional age / sex composition... 9 Figure 2.2. Tidal heights and numbers of sea lions onshore at Sunset Island... 15 Figure 3.1. Study site locations... 27 Figure 3.2. Steller sea lion counts before and after a research disturbance during summer... 29 Figure 3.3. Steller sea lion counts before and after a research disturbance during winter / spring...30 Figure 3.4. Coefficients of variation before and after a research disturbance... 30 Figure 3.5. Daily mean instantaneous rate of change... 34 Figure 3.6. Sea lion response to additional disturbance events... 36 Figure 3.7. Percentage of sea lions leaving haulouts in response to additional disturbances... 36 Figure 3.8. Steller sea lion response to disturbance... 39 Figure 4.1. Study site locations and proportional age / sex composition... 50 Figure 4.2. Numbers of focal animals hauling out of the water by age / sex class... 53 Figure 4.3. Kaplan-Meier survival curves: probability of sea lions remaining hauled out... 53 Figure 4.4. Numbers of behaviours per minute by age / sex class, season and before and after a research distubance... 55 Figure 4.5. Numbers of interactions per minute by age / sex class, season and before and after a research distubance... 56 vii

Acknowledgments I would like to thank Dr. Andrew Trites for his help and guidance throughout this journey and for providing me with such a fantastic project to work on. Thank you to my committee members Drs. A.R.E. Sinclair, D. Weary and A. Harestad for all of their help with the planning, analysis and preparation of my thesis. This project would not have been possible without funding support from the North Pacific Universities Marine Mammal Research Consortium, the Anne Vallée Ecological Fund and the Dean Fisher Memorial Scholarship in Zoology. Logistic and technical support provided by L. Jemison, J. Spicciani, B. Gisborne, G. Borgens, the Etzkorns, the Martins, Pacific Rim National Park Reserve, Parks Canada, Fisheries and Oceans Canada, Glacier Bay National Park, Alaska Department of Fish & Game, North Pacific Marine Science Foundation and the National Marine Fisheries Service, NOAA was greatly appreciated. I feel very fortunate to have worked with such amazing people within the Marine Mammal Research Unit, with special thanks to K. Soto, R. Joy, A. Winship, S. Kumagai, V. Deecke, D. Rosen, D. Tollit, K. Bodtker, T. Jeanniard du Dot, and P. Lestenkof for their comments and help with data analyses, R. Daniel, E. Bredesen, S. Ban, B. Wilson and M.A. Lea for their support and supplying data and P. Rosenbaum for logistical support. Thank you to all of my field and lab technicians M. Marcotte, E. Leung, J. Felio, A. Nayar, B. Smith, L. Wilson, D. Gummeson, M. Davies, L. Carneiro, R. Shao and of course, all of the Steller sea lions with whom we shared a bit of rock in the North Pacific. Thank you Michelle, Elaine and Lindsay, for your enthusiasm, spirit and warm memories of wonderful times in the field and Jon, for convincing me of your interest in sea lions, despite declining the job offer. Finally, from start to finish, it was assurance and confidence from my Mum, Brother and HM that encouraged me through this degree. This has been a dream of a project, taking me to remote, beautifully lush but not-so-tropical islands, and day after day, allowing me to watch such extraordinary animals on their own terms, in their own environment. viii

Chapter 1: General Introduction: Human disturbance and the hauling out behaviour of Steller sea lions Wildlife populations are affected to varying degrees by human disturbance (Allen et al., 1984; Lewis, 1987; Andersen et al., 1996; Fowler, 1999; Verhulst et al., 2001; Frid, 2003). Ranging from behavioural modifications to physiological stress responses, an animal s resting, feeding and breeding patterns may all be affected by human presence (Suryan and Harvey, 1999; Gill et al., 2001). Activity and development in remote areas has increased considerably in recent years raising concern about potential effects on animal populations. Since the late 1970 s, research on Steller sea lions (Eumetopias jubatus) has focused on a range of physiological, environmental and ecological factors to determine causes of an 85% population decline (Merrick et al., 1987; Trites and Larkin, 1996). Some of this research has resulted in repeated disturbance of sea lion haulouts and rookeries. The skittish nature of Steller sea lions usually results in entire colonies being disrupted when researchers go on shore to capture animals or to collect scats for dietary analyses. This raises behavioural and physiological concerns for populations experiencing repeated or high levels of human disturbance. Various methodologies used to evaluate the effects of disturbance on other species have often revealed species-specific responses (e.g., Salter, 1979; Born et al., 1999; Suryan and Harvey, 1999; Engelhard et al., 2002; Orsini, 2004). Anecdotal reports and limited knowledge of the effects of human disturbance on Steller sea lions necessitated a systematic approach to measuring sea lion response to disturbance. Understanding when and how animals use the environment is important for population conservation efforts (Reder et al., 2003), and determining response to research activities and subsequent recovery is an important step in determining the vulnerability of sea lions to disturbance. There are two main components in assessing response to disturbance. The first involves determining whether human presence causes animals to avoid disturbed areas, and the second requires long-term assessment of the physiological and distributional parameters of disturbed populations (Gill et al., 1996). Due to the lack of experimental data concerning Steller sea lion response to disturbance (see Lewis, 1987 for an assessment of a rookery disturbance), I chose 1

to address the first aspect by determining recovery (return of animals following human presence on the haulout) at haulouts and one rookery in an attempt to provide baseline data concerning the numbers of Steller sea lions that use haulouts following research activities. Ongoing long-term dietary studies of Steller sea lions using haulouts in southeast Alaska and British Columbia provided dates and locations of planned scat collection events. Knowing when researchers would be on shore collecting fecal samples provided a controlled disturbance, allowing me to experimentally measure the short-term effects of research activities on Steller sea lions by performing observations before, during and after scat collection. Observations were conducted at multiple locations with replicated sites in summer and winter / spring. Two control sites were included that were close to haulouts experiencing a research disturbance. The numbers of animals on land before the disturbance were compared to assess normal or baseline haulout use, and then examined for changes occurring as a result of the research disturbance. I addressed recovery both in terms of percentage of animals that returned following human presence on the haulout, and the time it took for sea lions to return in relation to hauling out trends. I also examined the daily variability in sea lion haulout use to more fully understand the factors affecting the numbers of sea lions on shore such as tidal height, time of day and the tidal stage (high tide, low tide). Such observations not only contribute to understanding the effects of disturbance on sea lion numbers, but they also have important implications in the interpretation of population surveys. The difference between the effects of disturbance on haulout populations (numbers of animals on shore) and individual sea lions (behaviour) is an important distinction. The first involved frequent counts to document recovery and determine if there were changes in numbers of animals on shore before and after the research disturbance. The second involved assessing individual behavioural changes by recording behaviours and interactions of sea lions that had just hauled out of the water. Pre- and post-disturbance behavioural follows of animals were compared to determine if research activities affected individual hauling out behaviour. Applying both methodologies provides greater insight into the effects that researchers have on the numbers of Steller sea lions using haulouts following human disturbance as well as the daily variability within and among sites. Documenting recovery rates may help to guide the issuing of research permits when considering the frequency of proposed fieldwork. 2

Overview The goals of my thesis were to determine the factors that influence the numbers of sea lions on shore, and the short-term effects of human disturbance on the hauling out numbers and behaviour of Steller sea lions during summer and winter / spring. My thesis has three principal chapters. Chapter 2 examines the influence of environmental and anthropogenic factors on the hourly and daily variability in numbers of sea lions hauled out at eight sites in southeast Alaska and British Columbia in summer and winter / spring. Chapter 3 considers the short-term effects of human disturbance on the numbers of Steller sea lions on land, and addresses three measures of local population recovery. Finally, Chapter 4 addresses the effects of season and disturbance on individual sea lion hauling out behaviour, specifically examining the likelihood of an animal remaining on land after hauling out, as well as subsequent behaviours and interactions with other animals. Chapters 2, 3 and 4 were written as independent manuscripts, and contain some redundancies because they were based on the same set of behavioural observations and counts of Steller sea lions made at multiple sites over two seasons. Field research was performed under Federal Marine Mammal permits: 358-1564 and 715-1457; and University of British Columbia Animal Care certificate: Category B, Protocol Number A02-0190. Literature cited Allen, S.G., D.G. Ainley, G.W. Page and C.A. Ribic. 1984. The effect of disturbance on harbor seal haul out patterns at Bolinas Lagoon, California. Fisheries Bulletin. 82:493-500. Andersen, R., J.D.C. Linnell and R. Langvatn. 1996. Short term behavioural and physiological response of moose Alces alces to military disturbance in Norway. Biological Conservation. 77:169-176. Born, E.W., F.F. Riget, R. Dietz and D. Andriashek. 1999. Escape responses of hauled out ringed seals (Phoca hispida) to aircraft disturbance. Polar Biology. 21:171-178. Engelhard, G.H., A.N.J. Baarspul, M. Broekman, J.C.S. Creuwels and P.J.H. Reijnders. 2002. Human disturbance, nursing behaviour, and lactational pup growth in a declining southern elephant seal (Mirounga leonina) population. Canadian Journal of Zoology. 80:1876-1886. Fowler, G.S. 1999. Behavioural and hormonal responses of Magellanic penguins (Spheniscus magellanicus) to tourism and nest site visitation. Biological Conservation. 90:143-149. 3

Frid, A. 2003. Dall's sheep responses to overflights by helicopter and fixed-wing aircraft. Biological Conservation. 110:387-399. Gill, J.A., K. Norris and W.J. Sutherland. 2001. Why behavioural responses may not reflect the population consequences of human disturbance. Biological Conservation. 97:265-268. Gill, J.A., W.J. Sutherland and A.R. Watkinson. 1996. A method to quantify the effects of human disturbance on animal populations. Journal of Applied Ecology. 33:786-792. Lewis, J.P. 1987. An evaluation of a census-related disturbance of Steller sea lions. Fairbanks, AK, University of Alaska: M.Sc. Thesis. 93 p. Merrick, R.L., T.R. Loughlin and D.G. Calkins. 1987. Decline in abundance of the northern sea lion, Eumetopias jubatus, in Alaska, 1956-86. Fisheries Bulletin. 85:351-365. Orsini, J.P. 2004. Human impacts on Australian sea lions (Neophoca cinerea) hauled out on Carnac Island (Perth, Western Australia), implications for wildlife and tourism management. Western Australia, Murdoch University: Ph.D. Thesis. 134 p. Reder, S., C. Lydersen, W. Arnold and K.M. Kovacs. 2003. Haulout behaviour of high Arctic harbour seals (Phoca vitulina vitulina) in Svalbard, Norway. Polar Biology. 27:6-16. Salter, R.E. 1979. Site utilization, activity budgets, and disturbance responses of Atlantic walruses during terrestrial haul-out. Canadian Journal of Zoology. 57:1169-1180. Suryan, R.M. and J.T. Harvey. 1999. Variability in reactions of Pacific Harbor seals, Phoca vitulina richardsi, to disturbance. Fisheries Bulletin. 97:332-339. Trites, A.W. and P.A. Larkin. 1996. Changes in the abundance of Steller sea lions (Eumetopias jubatus) in Alaska from 1956 to 1992: how many were there? Aquatic Mammals. 22:153-166. Verhulst, S., K. Oosterbeek and B.J. Ens. 2001. Experimental evidence for effects of human disturbance on foraging and parental care in oystercatchers. Biological Conservation. 101:375-380. 4

Chapter 2: Single census counts may not accurately reflect numbers and trends of Steller sea lions using haulout sites Summary Accurate estimates of pinniped population size are needed to properly monitor and manage seals and sea lions. Most pinniped populations are counted from aerial photos taken at a single point in time. Aerial censuses of pinnipeds are usually scheduled to ensure peak numbers on land and can sometimes be corrected for tidal effects, time of day and other environmental conditions. I sought to evaluate the reliability of single counts of Steller sea lions (Eumetopias jubatus) made at haulouts and at a single rookery. Numbers of Steller sea lions were counted at 20-minute intervals during summer (n = 7 sites) and winter / spring (n = 6 sites) in southeast Alaska and British Columbia. Maximum generalized linear models showed that tidal height, time of day, tidal stage (high tide, low tide), season and location affected sea lion haulout numbers. However, daily and seasonal hauling out patterns appeared to be site-specific for all sets of observations. Numbers of sea lions on shore varied considerably among hours on any given day, such that there was the potential to miss as many as 49% (mean; range 20 86%) of the sea lions that used a haulout on any given day. Coefficients of variation for hourly counts of sea lions ranged from 16 48% for individual haulouts (average CV = 32%), and was much lower at the one rookery that I examined (CV = 9%). Single counts of sea lions using haulouts were highly variable and could not be corrected for environmental influences on hauling out behaviour using a single correction factor. Consideration should be given to developing sitespecific correction factors that address the variety of environmental factors that affect the numbers of sea lions on land. In general, rookery counts have less variability than haulout counts and are likely to yield better indices of sea lion population sizes and trends. Introduction Most seals and sea lions haul out onto land or ice to rest between feeding trips, or to give birth, mate and engage in social interactions (Salter, 1979; Calambokidis et al., 1987; Watts, 1996; Reder et al., 2003; Orsini, 2004). Predator avoidance as well as thermoregulation may also play an important role in determining when and for how long individual animals remain out of 5

water (Daniel, 2003; Fernández-Juricic and Schroeder, 2003). Haulouts consist of groups of individuals that may vary by age and sex classes among locations. Overall, seals and sea lions tend to use traditional sites that are free of predators and presumably close to areas of optimum feeding. Steller sea lions (Eumetopias jubatus) are quite gregarious and generally haul out of the water in large groups on rocky offshore islands along the North Pacific rim (Loughlin, 2002). Numbers of animals using any given haulout can range from fewer than fifty, to thousands of individuals. They do not migrate, but exhibit seasonal movements between breeding (rookeries) and non-breeding (haulouts) sites (Sease and York, 2003). Haulout and rookery locations are well documented, and in some cases are known to have been consistently used for centuries (Sease and Gudmundson, 2002, Ban, 2005). Adult males defend territories at rookeries during late spring / early summer (May June) where females give birth to pups, and mating occurs. The breeding population disperses during late summer / early fall (August November) to join the rest of the population at haulouts for the remainder of the year. The criteria used by Steller sea lions to select terrestrial sites may be influenced by environmental or behavioural factors, or by prey availability around haulouts (Ban, 2005; Call and Loughlin, 2005). Environmental factors such as tidal heights and tidal stage, time of day, season, distance to the nearest alternative haulout, weather, risk of predation and available space on the haulout might affect a sea lion s decision to haul out (Kastelein and Weltz, 1991; Grellier et al., 1996; Porter, 1997; Moulton et al., 2002; Reder et al., 2003). Similarly, the age and sex of a sea lion may dictate behavioural needs to haul out during certain times of the year to moult, nurse or reproduce (Bengston and Cameron, 2004). Numbers of animals hauling out at any given location might also reflect daily or seasonal movements or seasonal behavioural variation (Thompson, 1989). Numbers of Steller sea lions that haul out vary daily and seasonally (Calkins et al., 1999) however, this variation has yet to be quantified. Understanding the factors that affect numbers of sea lions that haul out could improve the precision of population estimates and trends, which are essential components of conservation plans and provide a basis for management decisions. Current estimates of the Steller sea lion population are based largely on aerial surveys. Aircraft are flown over traditional sea lion rookeries and haulout sites at an approximate altitude of 150 200 m. Each site is surveyed with overlapping photographs (80 200 mm 6

zoom lenses) at 100 150 knots air speed. Once the surveys are completed, magnified images are projected and independent replicate counts of sea lions are made (Sease and Gudmundson, 2002). Aerial survey analyses currently employ correction factors to estimate the total populations size (Trites and Larkin, 1996; Sease and Gudmundson, 2002). The correction factors address the proportion of animals that may be away from the haulouts and rookeries at the time of the census, as well as account for the proportion of animals that may not be visible to the surveyor due to visibility or site topography (Sease and Gudmundson, 2002). Pinniped surveys are typically flown between 1000 and 1600 h in an attempt to photograph peak numbers of animals on land. However, hauling out patterns may vary significantly among species. For example, harbour seals (Phoca vitulina) tend to be tidally restricted with peak numbers generally occurring at low tides (Henry and Hammill, 2001; Reder et al., 2003). Steller sea lion pups are also predictably found on land during their first few weeks of life, offering a temporal window of obligate haulout use. Such predictability in hauling out patterns reduces the error in estimating the true numbers of pinnipeds using any particular site. However, not all species or age groups may use terrestrial sites in such a predictable manner. In this Chapter, I examine the influence of environmental and anthropogenic factors on the hourly and daily variation in numbers of Steller sea lions at haulout sites in southeast Alaska and British Columbia during summer and winter / spring. The study was conducted at one rookery and seven haulout sites, and yielded results that have bearing on the interpretation of census data used to assess the status of Steller sea lions throughout their range. Methods Data collection Study sites were chosen for accessibility, ease of observing sea lions without detection, and suitability for camping within hiking distance to the sea lions. Final selection of sites involved consultation with the National Marine Fisheries Service, the Alaska Department of Fish and Game, Fisheries and Oceans Canada, Glacier Bay National Park and Pacific Rim National Park Reserve. Sea lions at all study sites formed part of the eastern genetic stock of Steller sea lions (Bickham et al., 1996). 7

Counts of Steller sea lions on shore were conducted for 2 3 weeks at one rookery and six haulouts between May and August of 2003. Five of the haulouts were revisited between February and April of 2004 to repeat the observations, while the breeding site was replaced with a year round haulout (Fig. 2.1). The sites were typically exposed rocky outcrops and shores with steep intertidal areas. Systematic counts were made at 20-minute intervals, 12 hours a day during summer (0800 2000), and during daylight hours during winter and spring (range 7 12 hours per day). Sea lions were observed from fixed location blinds with the aid of binoculars and spotting scopes to avoid detection. Sea lions were classified by age and sex class (pup, juvenile, female, subadult male and adult male). Age and sex class percentages were arcsine transformed and differences in composition among non-breeding haulouts were evaluated with Chi-square tests for proportional data (Zar, 1996). An F-test was used to determine differences in variance in numbers of sea lions on land (coefficients of variation) among islands. My study was part of a larger study designed to determine the effects of human disturbance on Steller sea lions (see Chapter 3). Hence, a predetermined disturbance occurred halfway through the observation period at each site to collect fecal samples for dietary analysis (except at Lowrie Island, where pup branding occurred). These disturbances were included as factors in the analyses of hauling out patterns. Scats were not collected at one site in the summer of 2003 (Carmanah 2, control) and at two sites during the 2004 winter / spring field season (Carmanah Point and SW Brothers Island where significant disturbances were respectively caused by a vessel approach followed by human-sea lion contact, and by eagles feeding on a dead sea lion, which both led to sporadic use of the haulout by the sea lions). 8

1 2 3 4 Summer Winter 1. Bulls Pups Juveniles Females SE Alaska 3. Subadult males 5 British Columbia 4. P F 1. South Marble Island 2. Benjamin Island 3. Sunset Island 4. SW Brothers Island 5. Lowrie Island 6. McInnes Island 7. Carmanah Point 0 100 200 km 6 7 5. 7. 6. B S 2. Figure 2.1. Study site locations and proportional age and sex composition of six Steller sea lion haulouts and one rookery (#5). All islands were visited between May August 2003 and again between February April 2004 except for Lowrie Island, which was replaced by Benjamin Island, AK in 2004. South Marble (Jun 14 Jul 8, Feb 22 Mar 9), Benjamin (Mar 17 Mar 31), Sunset (May 18 Jun 5, Feb 23 Mar 9), SW Brothers (May 18 Jun 5, Mar 16 Mar 31) and Lowrie (Jun 15 Jul 8) Islands are located in southeast Alaska, USA. McInnes Island (Jul 13 Jul 24, Apr 7 Apr 22) and Carmanah Point (Jul 30 Aug 14, Apr 7 Apr 22) are located in British Columbia, Canada. Generalized linear models Following analyses of population trends by Calkins et al. (1999) and Small et al. (2003), a maximum likelihood generalized linear model (S-Plus 6.2) was used to address multivariate influences of environmental and anthropogenic factors on the numbers of sea lions hauled out. This model contained an overdispersion parameter allowing for non-normal Poisson distributed data to be examined in relation to tidal height, time of day, tidal stage (high tide, low tide), the research disturbance and interactions between these variables. 9

Tidal stage was included in the model as a quadratic term to account for the non-linear aspect of fluctuating tidal levels. Low tide was given an integer value of 0, with subsequent changes in tide height assigned values of 1 (20 minutes before and after low tide), 2 (40 minutes before and after low tide), 3, 4, 5, etc. Quadratic tidal values were then calculated as 1, 4, 9, 16, 25, etc. Two tide variables were included in the model to measure potentially different aspects of the influence of tides on the numbers of sea lions hauled out. Tidal height directly affects the available space for animals to haulout, whereas daily variation in the height of low tide can have substantially different effects on the amount of available haulout substrate, even at the same tidal stage (i.e., low tide) (Small et al., 2003). Interactions between the above variables assessed relationships between terms and were included to rule out any underlying correlation between variables. The generalized linear model was applied to each study location and season in keeping with the methodology employed by Calkins et al. (1999), who estimated population trajectories for separate sites under the assumption that the trajectories differed among sites. I wanted to identify differences in covariate influence on the response variable. I therefore employed stepwise deletion and Akaike s Information Criterion (AIC) to determine the best model fit for each site, with the expectation that different combinations of covariates would be retained in the various models due to physical and locational differences in haulout sites. Main effects were not considered for elimination from the final model if they were included in a significant interaction term. As has been done in other analyses of population trends (Calkins et al., 1999; Moulton et al., 2002; Small et al., 2003) the Chi-square goodness-of-fit test was used to assess the significance of observed differences in numbers of sea lions hauled out with each model term, and results were considered statistically significant based on α = 0.05. Residuals were examined for correlations that might have influenced significance levels. Results Differences among study locations included the average number of animals hauled out, age and sex composition and the influence of environmental variables on the hauling out behaviour of the sea lion communities. 10

Numbers and variation of sea lions on land There was a large range in the mean numbers of animals observed hauled out among study sites (e.g., 18 ± 0.69 SE at McInnes Island and 447 ± 1.66 at Lowrie Island before disturbances occurred). This wide range in numbers hauled out made it difficult to compare locations with regards to the strength or influence of various factors on the number of animals hauled out. Coefficients of variation (CV, used as a relative measure of variability in the number of sea lions hauling out each day) were significantly different among islands before and after research activities (CV, F 9 = 0.314, p < 0.001); (CV range 9.36% - 48.20% before and 19.05% - 85.82% after the research disturbance, excluding Carmanah Point where the postdisturbance CV exceeded 100%). Both time periods were examined separately to rule out potential variance due to disturbance. Using daily midday counts from the days that preceded the research disturbance resulted in a CV of 35.29% over an average of 9 days, compared to a CV of 29.90% when hourly variation in numbers was accounted for (Table 2.1). Thus, the CV associated with midday counts differed by 18.03% (relative difference) when compared to counts made throughout the day (absolute difference 5.39% / hourly variation 29.90% * 100). Numbers of sea lions observed varied by an average of 49% between minimum and maximum counts (all sites and days combined) (Table 2.2). Maximum differences in numbers of sea lions ranged from 20% at Lowrie Island (breeding site) to 86% at McInnes Island during summer. Age / sex composition Age and sex classes of sea lions differed significantly among study sites during the summer months (χ 2 6 = 478.7 for juveniles, 142.6 for females, 37.8 for subadult males, 51.7 for adult males, p < 0.001) (Fig. 2.1). However, only the proportion of juveniles differed significantly among islands during winter / spring (χ 2 5 = 45.10, p < 0.001). Adult female, subadult male and adult male proportions were not significantly different among islands during the winter / spring months (χ 2 5, p > 0.05). 11

Table 2.1. Coefficients of variation for all count data and midday counts. The estimated variation between midday counts resulted in an absolute difference of 5.39% and relative difference of 18.03% (5.39/29.90 * 100 = 18.03%) if single counts were used instead of multiple counts per day. Only data from days preceding the research disturbance were included to calculate these coefficients of variation to exclude variation due to human disturbance. Numbers of counts are shown in parentheses. Coefficient of variation Season Location All counts Midday counts Summer 2003 Sunset 35.43 (n = 281) 27.81 (n = 8) Brothers 28.52 (304) 31.06 (8) Lowrie 9.36 (633) 9.29 (17) South Marble 28.47 (694) 42.60 (19) McInnes 48.20 (157) 74.08 (5) Carmanah Point 37.13 (320) 54.35 (11) Carmanah 2 27.53 (193) 25.89 (6) Winter / Spring 2004 Sunset 43.93 (169) 38.46 (7) Benjamin 30.55 (205) 47.26 (7) South Marble 16.01 (189) 16.54 (8) McInnes 23.75 (282) 20.87 (9) All seasons Mean (n = 11) 29.90% 35.29% Table 2.2. Maximum percent difference in daily counts [(max min) / max] * 100. The mean minimum and maximum daily counts were used to calculate the potential percent difference in observed numbers of sea lions on shore. This methodology was chosen to report differences ranging from 0 100%. Only data from days preceding the research disturbance were included (mean: 49%; range: 20 86%). Mean daily count Season Location Minimum Maximum % Difference Summer 2003 Sunset 39 148 74 Brothers 173 306 43 Lowrie 393 493 20 South Marble 149 324 54 McInnes 4 29 86 Carmanah Point 71 134 47 Carmanah 2 68 135 50 Winter / Spring 2004 Sunset 130 238 45 Benjamin 125 255 51 South Marble 179 253 29 McInnes 156 247 37 12

Factors that affect numbers of sea lions on land Different combinations of tidal height, time of day, tidal stage, the research disturbance and interactions between these variables significantly influenced the numbers of animals on land at the study sites (Table 2.3). The influence of covariates on daily variation was significant and varied among sites and seasons, demonstrating the need to include covariates in statistical analyses to address daily variation in numbers of animals hauled out. Tidal height. Tidal height significantly influenced the numbers of sea lions on shore at Sunset, South Marble, Carmanah Point and Carmanah 2 in summer and at Sunset, SW Brothers and Benjamin Islands during winter / spring (χ 2, p < 0.05, see Table 2.3 for degrees of freedom). Negative correlation coefficients indicated that increasing tidal heights corresponded with a decrease in numbers of sea lions on shore. Counts and tidal heights at Sunset Island in May and June 2003 are illustrative of the sets of observations made at haulouts (Fig. 2.2, see Chapter 3 for counts made at the other 12 sites). Time of day. Time of day significantly influenced the numbers of sea lions on shore at McInnes Island during the summer study period (χ 2 422, p = 0.011), and at Carmanah Point during both summer (χ 2 630, p = 0.017) and winter / spring seasons (χ 2 603, p < 0.001). However, only weak trends were documented at other locations. Research disturbance. The research disturbance significantly influenced the numbers of animals on land at nine of the 10 study locations, as well as at a nearby haulout during summer (Summer: Sunset, SW Brothers, Lowrie, South Marble, McInnes, Carmanah Point and Carmanah 2; Winter / Spring: Sunset, South Marble and McInnes: χ 2, p < 0.05). In two cases, scat collection did not occur due to a severe boat harassment at Carmanah Point in April 2004, and because eagles frightened the sea lions while eating a dead animal at SW Brothers Island in March 2004. The research disturbance did not significantly influence the numbers of sea lions hauled out at Benjamin Island in March 2004 (χ 2 489, p = 0.80). Tidal stage. Tidal stage significantly influenced the numbers of sea lions on shore during summer at Carmanah 2 (χ 2 408, p < 0.001), and during winter / spring at Sunset Island (χ 2 394, p = 0.001), and during both seasons at SW Brothers Island (summer: χ 2 696, p = 0.039, winter / spring: χ 2 516, p = 0.004) and Carmanah Point (summer: χ 2 630, p < 0.001; winter / spring: χ 2 603, p = 0.012). 13

Table 2.3. Factors affecting numbers of sea lions hauled out in summer (S) and winter / spring (W) months. Correlation coefficients for generalized linear models fit to the response variable of sea lion haulout count data at each study location. Stepwise deletions were performed to determine the best model fit for each study period. Insignificant covariates were included in the model if they appeared in a significant interaction term, but were deleted from the model otherwise. All models were significantly different than the null model. Significant covariates are shown in bold. There were no experimental disturbances at SW Brothers or Carmanah Point during winter / spring, and consequently BA and BA interactions were not included in these initial model equations. Carmanah 2 was not disturbed for scat collection, but was close to the Carmanah Point study location. The research disturbance factor was included in this model to account for the possibility of scat collection affecting more than one site in an area, specifically addressing the movement of animals between these two sites. Sunset Brothers Lowrie South Marble McInnes Carmanah Point Carm 2 Benjamin variables S W S W S S W S W S W S W tidal height -0.6-0.67-0.71-0.63-0.31-0.82-0.83 time -0.96-0.95-0.65-0.92 BA -0.6-0.57-0.49 na -0.84-0.69-0.66-0.17-0.71-0.62 Na -0.36 tide:time 0.11 0.35-0.63 0.04 0.14-0.23 0.52 tide:ba 0.77 BA:stage na 0.48 Na 0.43 na -0.21 Na dispersion parameter 0.12 0.52 0.06 1.17 0.021 0.09 0.27 0.44 0.06 1.82 1.64 0.12 0.13 df 704 394 696 516 894 959 432 422 552 630 603 408 489 Interactions between variables. The interaction between tidal height and the time of day showed that the effect of tidal height on sea lion numbers varied at different times of the day during the winter / spring study period at SW Brothers Island (χ 2 516, p < 0.001). The interaction between tidal height and the research disturbance indicated that the disturbance affected sea lions numbers differently at varying tidal heights during the summer at Carmanah Point (χ 2 630, p = 0.001). Similarly, the interaction between the research disturbance and tidal stage indicated that the effect of the disturbance differed depending on tidal stage during summer at Sunset Island and Carmanah Point (Sunset: χ 2 704, p = 0.007; Carmanah Point: χ 2 630, p < 0.001). 14

Numbers of sea lions on shore 200 150 100 50 0 480 325 170 15-140 Tidal height (cm) May 18 Consecutive observation days June 5 Figure 2.2. Tidal heights (lighter dashed lines) and numbers of sea lions on shore at Sunset Island from May 18 - June 5, 2003 (dark lines). These data are representative of other study sites visited, and show that numbers of sea lions on shore typically decrease as tidal height increases. Discussion Various environmental influences affected the hauling out behaviour of Steller sea lions. The combination and degree of influence that these factors had on the numbers of sea lions hauled out varied seasonally and by site. My study suggests that sea lions are subject to site-specific environmental effects, but that they alter their hauling out behaviour to accommodate various influences. These findings support those of other studies indicating that environmental influences have strong effects on the hauling out behaviour of pinnipeds (Calkins et al., 1999; Moulton et al., 2002; Small et al., 2003). Data obtained with systematic counts are valuable when assessing the reliability of aerial census counts in terms of optimal conditions in which to survey. Observations from specific haulouts and rookery locations may provide guidance for estimating daily variation in the numbers of animals hauled out at any given time. Ultimately, it is important to determine if these environmental covariates have a discernible effect on estimates of population trends. 15

Numbers of sea lions on land The numbers of animals hauled out differed significantly among study locations, but were assumed to reflect the natural distribution of animals at their preferred locations. Daily movements to and from haulouts accounted for some of the daily fluctuation in numbers of animals hauled out. Preference for haulout locations may be influenced by local prey abundance, shelter or occurrence along seasonal routes (Härkönen et al., 1999). Individual sea lions may prefer certain locations, returning regularly to particular haulouts, while other individuals may exhibit transient behaviours and move from one haulout to another (Raum- Suryan and Pitcher, 2002). These differences in site fidelity may account for some of the residual variance that was not explained by the generalized linear regression models. Topographic and age and sex differences among study locations may influence seasonal use of haulouts and may indicate potentially different types of haulouts. There may be inherent differences in hauling out behaviour that determine the numbers or proportions of certain age or sex classes on a haulout at any given time, and specifically over seasons (Reder et al., 2003; Sease and York, 2003). Seasonal changes in the behaviour of breeding and non-breeding individuals will influence site choice, while the distance to other available haulouts might affect the degree of local movement and flux among haulouts. Seasonal changes likely did not occur within the short study periods (2 3 weeks). However, research disturbances that cause animals to leave the haulout and occur within a few days of aerial surveys will directly influence the results of census efforts (Chapter 3). Environmental influences Environmental influences on the numbers of sea lions hauled out are substantial and biologically significant (Small et al., 2003). For that reason, determining influences on the numbers of sea lions on shore is extremely important for evaluating methods to assess population trends, especially in terms of conditions during population surveys. Most studies attempting to identify variables that affect pinniped hauling out trends have been done at single locations (Ono et al., 1987; Grellier et al., 1996; Porter, 1997; Moulton et al., 2002). However, it is clear from my assessment of multiple sites that seasonal and site-specific factors affect the numbers of sea lions on land. While in practice it is nearly impossible to conduct all aerial surveys and counts at optimal times and tidal stages, it is necessary to address these 16

influences on numbers of observed animals. Of added importance are the various influences that these factors have among locations. My study documented changes in numbers of sea lions using the haulouts in response to tidal height, time of day, tidal stage, a research disturbance and interactions between these variables. Time of day has typically been examined to determine circadian trends in hauling out patterns (Porter, 1997; Calkins et al., 1999). This may be related to available sunlight or prey availability in the water column. However, graphical analyses of individual study sites indicated only weak trends. The numbers of animals on land seemed to follow the tidal cycles more closely than the time of day. Numbers of sea lions hauled out differed significantly at nine of ten sites that experienced a research disturbance, as did numbers at a nearby undisturbed haulout (Carmanah 2) (see Chapter 3 for further details concerning the research disturbance). It is difficult to quantify weather variables in terms of intensity and duration, or to document subjective qualitative descriptions of conditions (i.e., intense snowfall, light rain or downpours), whereas tidal stages, time of day and human presence all have discreet measurements at distinct intervals. Other weather related factors such as temperature, wave and water state, precipitation, wind and solar radiation should be included in future analyses. The degree to which tide influences hauling out behaviour is likely related to the physical geography of a site (Kastelein and Weltz, 1991). Tidal height significantly influenced the numbers of animals hauled out during at least one season at seven of eight study locations. The eighth location, Lowrie Island, was a breeding site. Steller sea lion rookeries appear to have limited tidal disturbance, which may enhance the safety of pups during their first few weeks of life. Correlation coefficients revealed a negative correlation among increasing tidal levels and sea lion numbers, as noted in other pinniped studies (Allen et al., 1984; Porter, 1997; Henry and Hammill, 2001). My observations indicated that preferred exposed haulout areas can be a limiting factor in numbers of animals on shore. Sea lions returning from trips to sea might be restricted from hauling out due to increasing tidal levels, and other animals may be forced to leave the haulout due to submersion. Sea lion foraging behaviour may correspond to tidal stage (high tide, low tide) and the effects of incoming and outgoing water that may influence prey behaviour around the haulout (Baumgartner, 1997). Tidal variation affects the numbers of animals visible during aerial surveys, which in turn affects the evaluation of 17

changes in abundance and distribution of animals. My results support the general conclusions of other studies indicating that environmental variables have varying influences on the hauling out behaviour of pinnipeds (Small et al., 2003). Estimated variation Steller sea lion hauling out behaviour is an important component of interpreting census counts and formulating population trend estimates. Correction factors calculated for the proportion of animals that may be away from land at the time of a census count attempt to account for the high mobility of sea lions. Aerial surveys typically replicate counts over subsequent days or years to calculate coefficients of variation for observed numbers of sea lions on shore (Sease and Gudmundson, 2002). One of the most difficult aspects of calculating correction factors for animals that may be away from haulout locations is estimating their variance. Replicating counts at individual sites allows for coefficients of variation (CV) to be calculated as a measure of variability in observed numbers of sea lions hauled out. Although this measure of variation may account for some of the environmental influences that affect the numbers of sea lions hauled out, on its own it cannot pinpoint the factors that affect the numbers of sea lions on land. Reported CV s for pinniped aerial surveys range from 2.3 12% for harbour seals, and up to 48 139% for Pacific walrus (Odobenus rosmarus) (Ferrero et al., 2000; Jeffries et al., 2003). Assessment of multiple Steller sea lion pup aerial surveys (1997, 1998) resulted in CVs ranging from 1 to 16.22% (Snyder and Pitcher, 2001). In contrast, replicated sites during the 2002 Steller sea lion population census of rookeries resulted in an extremely low coefficient of variation of 2.9% (Sease and Gudmundson, 2002). To assess whether the use of replicated counts is consistent with the variation that was observed, I calculated the difference between CV s for midday counts (a single measure per day) and CV s for all counts made during my study period (Table 2.1). CV s calculated for midday counts resulted in an average CV of 35.29% (range 9.29 74.08%). Using all counts from days preceding the research disturbance resulted in an estimated average CV of 29.9% for all haulouts (range 16.0 48.2%). Such values indicate that a single census count might not reasonably reflect the numbers of sea lions using the haulout on the day of the survey. The 18